Detecting adaptation involves comparing the performance of populations evolving in different environments. This detection may be confounded by effects due to the environment experienced by organisms prior to the test. We tested whether such confounding effects occur, using spider-mite selection lines on two novel hosts and one ancestral host, after 15 generations of selection. Mites were either sampled directly from the selection lines, or subjected to a common juvenile or to a common maternal environment, mimicking the most frequent environmental manipulations. These environments strongly affected all life-history traits. Moreover, the detection of adaptation and of correlated responses on the ancestral host was inconsistent among environments in almost 20% of the cases. Indeed, we did not detect responses unambiguously for any life-history trait. This inconsistency was due to differential environmental effects on lines from different selection regimes. Therefore the detection of adaptation requires a careful control of these environmental effects.